The xML Lab at the National University of Singapore has introduced the Jeva model tailored for decision-making needs. This model addresses four major issues found in traditional models: susceptibility to judgment alteration due to option order changes, forced output when no suitable option is available, limited scalability in candidate options, and redundant repetitive computations. The Jeva model incorporates four core architectural features: null awareness, permutation invariance, option scalability, and cacheability. It achieves relevant capabilities by separating text understanding from judgment, involving the full set of candidates in preference computation, distinguishing between relative candidate preferences and answer existence, and supporting the caching and reuse of candidate semantic representations. Experimental results demonstrate that Jeva excels in various aspects, including empty answer detection, judgment stability under option rearrangement, accuracy and speed with large candidate sets, and reasoning efficiency during candidate reuse, thereby providing a novel design approach for language model decision-making tasks.